I’m a first-year machine learning PhD student, and I’m wondering how best to spend my PhD to prepare for policy positions (as a US citizen, I’m especially looking at programs like TechCongress and AAAS). What skills should I develop, and what can I do to develop them? What topic areas should I become an expert in? Should I learn about subjects broadly or just zero in on my PhD topic? I’m also wondering how much overlap there is between work that would best improve my resume for policy and work that would increase my chances of landing in academia. Roughly, there’s a spectrum of how to allocate my PhD resources with the following two extremes: on one, I can try to pursue traditional academic success at all costs (publish a lot, network heavily with academics); on the other, use the PhD funding to subsidize my work in other areas (e.g. policy research) and just do the bare minimum required to graduate by thesis.
More on my background/situation. My current PhD topic is fairness. I’m not particularly interested in value alignment or X-risk AI problems; I also don’t feel like I’m well-equipped to research those topics. I’m in a UK program and so my PhD is quite a bit shorter, and I will graduate (or my funding runs out) in 2023⁄24. I have a generalist background/disposition, and I’m fortunate to be part of a team that also includes philosophers (lots of them), lawyers, and sociologists to stimulate those interests. My university used to have a better reputation for ML, but now it’s declined a bit. It’s still very well-regarded in philosophy. Given that situation, becoming traditionally well-regarded in just ML is going to be very difficult and unlikely but not impossible. Thus far, I’ve found that my interest in doing that has been flagging, since I worry a lot about the immediate implications of a lot of ML research on society. But I think I could buckle down if I found the right reasons for doing so.
Thanks for doing this, especially since my last comment on the 80K open thread went unanswered (no shade though! I think I wasn’t specific enough, which I’ll try to do here). In that comment, I expressed a lot of uncertainty around: How good do I have to be at academic research, in order to do a PhD for policy? Since then, through informational interviews, I’ve gotten the picture that merely satisficing at academic research should be enough, but I shouldn’t treat it as a trivial constraint—doing good science is still hard, and at the end of the day, technology policy still requires some amount of expertise. I’d still appreciate thoughts on this question though, if you have some!
In addition to that question, I’ve begun to develop a new question: What should I do if none of my top options work out (if I fail or decide that neither academia nor policy are for me?) Throughout university and my career (I do part-time curriculum writing for an education startup), I’ve consistently gotten the feedback that I have a knack for teaching: for explaining concepts intuitively for others, for scaffolding ideas so that they make sense to others, and anticipating common learning obstacles. The problem is that there doesn’t seem to be an EA-aligned career path in which these skills would be directly relevant. 80K recommends against teaching, and I mostly agree with their assessment. Are there any EA opportunities in which the skills I’ve listed above would be relevant?